Re: [PATCH] linux-user: Correct definition of stack_t

[Peter Maydell]

On Thu, 5 Nov 2020 at 15:48, LemonBoy <thatlemon@gmail.com> wrote:
> On 05/11/20 16:16, Peter Maydell wrote:
> > Coincidentally, I just noticed this yesterday (am in the middle of
> > trying to debug why qemu-sparc64 crashes trying to start bash...)
> >
>
> What an interesting coincidence, I also stumbled across this bug while
> debugging a crash with some sparc64 binary.

Don't suppose it involves sigsetjmp/siglongjmp?
So far I have discovered that the linux-user/sparc
target_ucontext and friends have host pointers where
they should have abi_ulong and also that target_mc_fpu
needs explicit 16-alignment (patch written, will post it
sometime this week), and once past that have hit another
bug that seems to be down to the setcontext/getcontext
traps not saving and restoring all the right registers.

thanks
-- PMM